"CECR-Laplacian-Pyramid Methane Concentration Dataset"

"This study proposes a novel gas concentration prediction method based on adaptive decomposition and multiscale feature modeling. The original methane concentration time series is preprocessed with wavelet denoising and subsequently decomposed using CEEMDAN into intrinsic mode functions (IMFs). Fuzzy entropy quantifies IMF complexity, enabling clustering-based reconstruction to separate components of differing complexities. Further, Laplacian pyramid decomposition extracts multiscale features from reconstructed signals, capturing distinct frequency characteristics. The resulting features are utilized to train predictive models for both single-step and multi-step forecasting of methane concentration in underground coal mines. Experimental results demonstrate significant improvements in prediction accuracy and robustness compared to baseline methods. This approach offers a promising framework for environmental gas monitoring and early warning systems."